photometric properties of void galaxies in the …photometric properties of void galaxies in the...

PHOTOMETRIC PROPERTIES OF VOID GALAXIES IN THE SLOAN DIGITAL SKY SURVEY

Randall R. Rojas,1Michael S. Vogeley,

1Fiona Hoyle,

1and Jon Brinkmann

2

Receivved 2004 July 2; accepted 2004 August 18

ABSTRACT

Using a nearest neighbor analysis, we construct a sample of void galaxies from the Sloan Digital Sky Survey(SDSS) and compare the photometric properties of these galaxies to the population of nonvoid (wall ) galaxies. Wetrace the density field of galaxies using a volume-limited sample with zmax ¼ 0:089. Galaxies from the flux-limitedSDSS with z � zmax and fewer than three volume-limited neighbors within 7 h�1 Mpc are classified as void gal-axies. This criterion implies a density contrast ��=� < �0:6 around void galaxies. From 155,000 galaxies, weobtain a subsample of 13,742 galaxies with z � zmax , from which we identify 1010 galaxies as void galaxies. Toidentify an additional 194 faint void galaxies from the SDSS in the nearby universe, r P 72 h�1 Mpc, we employvolume-limited samples extracted from the Updated Zwicky Catalog and the Southern Sky Redshift Survey withzmax ¼ 0:025 to trace the galaxy distribution. Our void galaxies span a range of absolute magnitude from Mr ¼�13:5 to �22.5. Using SDSS photometry, we compare the colors, concentration indices, and Sersic indices of thevoid and wall samples. Void galaxies are significantly bluer than galaxies lying at higher density. The populationof void galaxies with Mr P M � þ 1 and brighter is on average bluer and more concentrated ( later type) thangalaxies outside of voids. The latter behavior is only partly explained by the paucity of luminous red galaxies invoids. These results generally agree with the predictions of semianalytic models for galaxy formation in cold darkmatter models, which indicate that void galaxies should be relatively bluer, more disklike, and have higher specificstar formation rates.

Subject headinggs: cosmology: observations — galaxies: photometry — galaxies: structure —large-scale structure of universe — methods: statistical

1. INTRODUCTION

Since the discovery of the void in Bootes (Kirshner et al.1981), with a diameter of 50 h�1 Mpc, and subsequent discov-eries of voids in larger redshift surveys (Geller & Huchra1989; Pellegrini et al. 1989; da Costa et al. 1994; Shectman et al.1996; El-Ad et al. 1996, 1997; Muller et al. 2000; Plionis &Basilakos 2002; Hoyle & Vogeley 2002), these structures haveposed an observational and theoretical challenge. Because thecharacteristic scale of large voids was comparable to the depthof early redshift surveys, few independent structures were de-tected, making statistical analysis of their properties difficult.Likewise, the limitations of computing technology constrainedearly cosmological simulations to include only a few voids persimulation.

Whether voids are empty or not has been the question of re-cent debate. Peebles (2001) pointed out the apparent discrep-ancy between cold darkmatter (CDM)models and observations.CDM models predict mass, and hence possibly galaxies insidethe voids (Dekel & Silk 1986; Hoffman et al. 1992). However,pointed observations toward void regions failed to detect a sig-nificant population of faint galaxies (Kuhn et al. 1997; Popescuet al. 1997; McLin et al. 2002). Surveys of dwarf galaxies in-dicate that they trace the same overall structures as larger gal-axies (Bingelli 1989). Thuan et al. (1987), Babul & Postman(1990), and Mo et al. (1994) showed that galaxies had commonvoids regardless of Hubble type.

Grogin & Geller (1999, 2000) identified a sample of 149 gal-axies that lie in voids traced by the Center for Astrophysics

(CfA) survey. The void galaxies were found in the Century and15R redshift samples. Grogin & Geller showed that the voidgalaxies tended to be bluer and that a significant fraction ofthem were of late type. Their sample of 149 void galaxies cov-ered a narrow range of absolute magnitude (�20 P B P� 17)of which 49 have a low-density contrast of ��=� ��0:5.Here we present a sample of �103 void galaxies found inregions of density contrast ��=� � �0:6. This sample is largeenough to allow comparison of void and wall galaxies with thesame color, surface brightness profile, and luminosity to statis-tically quantify their differences. The range of absolute mag-nitude (Sloan Digital Sky Survey [SDSS] r band) in our sample(�22 P Mr P 13) is large enough to include faint dwarfs togiants.In this paper, we introduce a new sample of void galaxies

from the SDSS. The large sky coverage and depth of the SDSSprovides us with the opportunity to identify for the first timemore than 103 void galaxies with ��=� � �0:6. In x 2 we dis-cuss the galaxy redshift samples that we use for this anal-ysis. In x 3 we describe our method for finding void galaxies. Inx 4 we present the results found from the comparison ofthe photometric properties of void and wall galaxies, and in x 5we interpret these results by comparing them to predictionsfrom semianalytic modeling of structure formation and prop-erties of different galaxy types. Finally, in x 6 we present ourconclusions.

2. THE REDSHIFT SURVEYS

The search for void galaxies requires a large three-dimensional map of the galaxy density field. We extract avolume-limited sample from the SDSS data to map the galaxydensity field and look for void galaxies in the full magnitude-limited sample. As the SDSS currently has a slicelike geometry,

1 Department of Physics, Drexel University, 3141 Chestnut Street, Phila-delphia, PA 19104; [email protected], [email protected],[email protected].

2 Apache Point Observatory, P.O. Box 59, Sunspot, NM 88349-0059.

50

The Astrophysical Journal, 617:50–63, 2004 December 10

# 2004. The American Astronomical Society. All rights reserved. Printed in U.S.A.

with each slice only �6� thick, large voids of radius �10 h�1

Mpc (h � H0=100 km s�1 Mpc�1) can be detected only atcomoving distances of rk 102 h�1 Mpc using the SDSS dataalone. Therefore, to trace the local voids, we also extract avolume-limited sample from the combined Updated ZwickyCatalog (UZC; Falco et al. 1999) and Southern Sky RedshiftSurvey (SSRS2; da Costa et al. 1998). It should be noted thatnearby void galaxies are not selected from the UZC and SSRS2surveys. These surveys are only used to define the density fieldaround SDSS galaxies that lie at distances r � 72 h�1 Mpc.

To recap, we have two volume-limited samples, one fromthe SDSS and one from the combined UZC+SSRS2. These sam-ples are used to define the galaxy density field only. Void gal-axies are found from the magnitude-limited SDSS sample. Wedefine the distant sample to be the SDSS magnitude-limitedsample truncated at 100 h�1 Mpc � r � 260 h�1 Mpc. Thenearby sample is the SDSS magnitude-limited sample truncatedat r ¼ 72 h�1 Mpc. Both magnitude-limited samples (nearbyand distant) are constructed using the SDSS r band. In thissection we describe each of the surveys and samples in detail.

2.1. The SDSS

The SDSS is a wide-field photometric and spectro-scopic survey. The completed survey will cover approximately104 deg2. CCD imaging of 108 galaxies in five colors andfollow-up spectroscopy of 106 galaxies with r < 17:77 will beobtained. York et al. (2000) provides an overview of the SDSS,and Stoughton et al. (2002) describes the early data release(EDR) and details of the photometric and spectroscopic mea-surements made from the data. Abazajian et al. (2003) describesthe first data release (DR1) of the SDSS. Technical articles pro-viding details of the SDSS include descriptions of the photo-metric camera (Gunn et al. 1998), photometric analysis (Luptonet al. 1999), photometric system (Fukugita et al. 1996; Smithet al. 2002), photometric monitor (Hogg et al. 2001), astrome-tric calibration (Pier et al. 2003), selection of the galaxy spectro-scopic samples (Strauss et al. 2002; Eisenstein et al. 2001), andspectroscopic tiling (Blanton et al. 2003a). A thorough analysisof possible systematic uncertainties in the galaxy samples isdescribed in Scranton et al. (2002).

We examine a sample of 155,126 SDSS galaxies (Blantonet al. 2003b; SAMPLE10) that have both completed imagingand spectroscopy. The area observed by SAMPLE10 is ap-proximately 1.5 times that of the DR1 (Abazajian et al. 2003).To a good approximation, the sample that we analyze con-sists of roughly three regions covering a total angular area of1986 deg2. Because of the complicated geometry of the SDSSsky coverage, the survey regions are best described in the SDSScoordinate system (see Stoughton et al. 2002). Where possiblein this section we describe approximate limits in the more fa-miliar equatorial coordinates. The first region is an equatorialstripe in the north Galactic cap (NGC). This stripe has a max-imum extent of 7N5 in the declination direction over the R.A.range 21h30m P � P 4h10m and a maximum length of 120

�

over the R.A. range 8h P � P 16h. The second region is in thesouth Galactic cap (SGC). There are three stripes, the bound-aries of which are defined in the SDSS coordinate system. Eachstripe is 2N5 wide in SDSS survey coordinates. One stripe iscentered at � ¼ 0� and covers the R.A. range 20h40m P � P2h20m. The other two stripes are above and below the equatorand cover similar R.A. ranges. In survey coordinates these twostripes cover the range �28� P k P 41�, 130� P � P 135� and�57� P k P 58�, 155� P � P 160�. The third large region is inthe NGC. In the SDSS survey coordinates it covers the range

�48� P k P 50�, 75� P � P 85�. There are additional smallerstripes at �37

� P k P � 22�, 60

� P � P 70�and 2

� P k P60�, 90� P � P 100� (the boundary is an approximation be-cause of the tiling geometry).

We correct the velocities of galaxies to the Local Group (LG)frame according to

�v ¼Vapex½sin(b) sin (bapex)þ cos (b)cos (bapex)cos (l� lapex)�;ð1Þ

where lapex ¼ 93�, bapex ¼ �4�, and Vapex ¼ 316 km s�1

(Karachentsev & Makarov 1996). The magnitudes of the gal-axies are K-corrected as described in Blanton et al. (2003c), andcorrections for Galactic extinction are made using the Schlegelet al. (1998) dust maps. Finally, to convert redshifts into comov-ing distances we adopt an (�m; ��) ¼ (0:3; 0:7) cosmology.

The decrease of observed galaxy density with distance in anapparent magnitude–limited galaxy sample might cause us toerroneously detect more voids at large distances. Therefore, weuse a volume-limited subsample of the SDSS to define thedensity field of galaxies. This sample consists of galaxies withredshifts less than the redshift limit, zmax, and SDSS r-bandabsolute magnitudes brighter than Mcrit , where

Mcrit ¼ r lim � 25� 5 log10½dl(zmax)�; ð2Þ

r lim3 is the magnitude limit of the survey, and dl is the luminosity

distance in units of h�1 Mpc at zmax. We form a volume-limitedsample of the SDSS with zmax ¼ 0:089, with a correspondingabsolute magnitude limit Mcrit ¼ �19:87 (in the SDSS r band).The redshift limit zmax ¼ 0:089 allows us to construct the largestpossible volume-limited sample from the current SDSS sample.This volume-limited sample contains 22,866 galaxies, where themean separation between these galaxies is �5.3 h�1 Mpc. For a(�m ¼ 0:3; �� ¼ 0:7) cosmology, the redshift limit of zmax ¼0:089 corresponds to a comoving distance of 260 h�1 Mpc.The lower bound of 100 h�1 Mpc on the comoving distance isnecessary because of the slicelike geometry of the early SDSSslices. Recall that voids of diameter k10 h�1 Mpc can only befound at rk100 h�1 Mpc, as discussed in x 2.

2.2. The Updated Zwicky Catalogg

The UZC (Falco et al. 1999) includes a reanalysis of datataken from the Zwicky Catalog and CfA surveys (Zwicky et al.1961, 1965; Zwicky & Herzog 1962–1965; Zwicky & Kowal1968; Geller & Huchra 1989; Huchra et al. 1990, 1995, 1999)together with new spectroscopic redshifts for some galaxies andcoordinates from the digitized POSS II plates. Improvementsover the previous catalogs include estimates of the accuracy ofthe CfA redshifts and uniformly accurate coordinates at the lessthan 200 level.

The UZC contains 19,369 galaxies. Of the objects with limit-ing apparent magnitude mZw � 15:5, 96% have measured red-shifts, giving a total number of 18,633 objects. The catalogcovers two main survey regions: 8h < �1950 < 17h, 8N5 <�1950 < 44N5 in the NGC and 20h < �1950 < 4h, �2N5 <�1950 < 48� in the SGC. We correct the velocities of the galax-ies with respect to the LG as discussed in x 2.1. The magni-tudes of the galaxies are corrected for Galactic extinction usingthe Schlegel et al. (1998) dust maps, and the magnitudes are

3 We use r lim ¼ 17:5 instead of r lim ¼ 17:77 in the construction of thevolume-limited catalog to ensure that we have a uniform limit across all thedata, since earlier stripes were only observed to r lim ¼ 17:5.

PHOTOMETRIC PROPERTIES OF SDSS VOID GALAXIES 51

K-corrected assuming K ¼ 3, which is appropriate for the Bfilter and the median galaxy morphological type Sab (Park et al.1994; Pence 1976; Efstathiou et al. 1988).

We construct a volume-limited UZC sample with zmax ¼0:025, since this is the redshift at which the largest volume-limited sample can be obtained. This volume-limited samplecontains 4924 galaxies and has a comoving depth of �76 h�1

Mpc and absolute magnitude limit ofMlim � �18:96 (BZw). Tocompare this limit to that of the SDSS, we translate a B-bandmagnitude into an approximate r-band magnitude of Mr ¼�20:06 using g� r ¼ 0:66 and g� B ¼ �0:45 from Fukugitaet al. (1996). The absolute magnitude limit of the UZC sampleis therefore slightly brighter than the SDSS limit. To ensure thatthis sample and the SSRS2 described below are equally deep,we cut back this sample to 72 h�1 Mpc.

2.3. The SSRS2

The SSRS2 galaxy sample (da Costa et al. 1998) was selectedfrom the list of nonstellar objects in theHubble Space TelescopeGuide Star Catalog. The SSRS2 contains 3489 galaxies in theSGC over the angular region: �1950 � �2N5 and b � �40�, cov-ering a total of 1.13 sr with mSSRS2 � 15:5, where the zero-point offset from the Zwicky magnitude system used in theUZC is approximatelymSSRS2 � mZw � 0:10mag (Alonso et al.1994).

We construct a volume-limited sample with the same redshiftlimit as for the UZC, zmax ¼ 0:025 (same reason as discussed inx 2.2) and (after adjustment of the zero point)M ��18:96. Forour chosen cosmology, the depth of the sample is�73 h�1Mpc,which we also cut back to 72 h�1 Mpc as discussed in the casefor the UZC sample. Therefore, both SSRS2 and UZC volume-limited samples have the same comoving depth.

As above (x 2.1), we correct galaxy velocities to the LGframe, apply the Galactic dust corrections based on the Schlegelet al. (1998) dust maps, and assume K ¼ 3 to K-correct the

observedmagnitudes. This volume-limited sample includes 725galaxies.The SSRS2 provides angular coverage in the SGC. The

combined UZC+SSRS2 sample contains 5649 galaxies and skycoverage of �4.25 sr.

2.4. Summary of Survveys

Figure 1 (left) shows an Aitoff projection of the three sur-veys. The black sections show the SDSS galaxies, and the dotsshow the UZC+SSRS2 galaxies. This figure demonstrates thatin terms of area, the SDSS is almost totally embedded in theUZC+SSRS2 data apart from along the bottom edge of thenorthern equatorial slice and a small part of the southernmostslice. Therefore, the combined UZC+SSRS2 survey is usefulfor defining the large-scale galaxy density field around theSDSS sample out to a distance of approximately 72 h�1 Mpc.Figure 1 (right) shows a cone diagram of the SDSS data with

j�jP15�. The inner circle is drawn at 72 h�1 Mpc, which is thecomoving depth of the combined UZC and SSRS2 volume-limited samples. The outer circle is drawn at 260 h�1 Mpc,which is the comoving depth of the SDSS volume-limited sam-ple. Beyond 72 h�1 Mpc, the selection function (the numberof observed galaxy density with distance) of these shallowersurveys drops and the thickness (in the declination direction)of the SDSS itself is adequate to define the density field aroundthe SDSS galaxies.

3. IDENTIFYING VOID GALAXIES USINGTHE NEAREST NEIGHBOR STATISTICS

We search for void galaxies in the SDSS using the nearestneighbor statistic. The two volume-limited samples (SDSSand UZC+SSRS2) are used to trace the voids. Any magnitude-limited galaxy that lies away from the boundary of the volume-limited sample and has less than three volume-limited sample

Fig. 1.—Left: Aitoff projection in celestial coordinates of the current SDSS galaxy redshift survey data (black sections) and the combined UZC and SSRS (dots).Approximate coordinates of each region are given in the text in x 2. Right: Cone diagram of the flux-limited SDSS data with j�j P 15�. The inner circle is drawn at72 h�1 Mpc, which is the depth of the combined UZC and SSRS2 volume-limited sample. The outer circle is drawn at 260 h�1 Mpc, which is the depth of the SDSSvolume-limited sample. In the region 100 h�1 Mpc < r < 260 h�1 Mpc we use the SDSS data to trace the distribution of the voids. However, nearby the SDSS iscurrently limited in volume, as only narrow strips have been observed. Therefore, nearby we use the UZC+SSRS2 to trace the voids out to 72 h�1 Mpc.

ROJAS ET AL.52 Vol. 617

neighbors in a sphere of 7 h�1 Mpc is considered a void galaxy.We expand on each of these steps below.

3.1. Proximity to Survvey Boundary

Galaxies in the magnitude-limited SDSS samples that lienear the boundaries of the volume-limited samples have sys-tematically larger distances to their third nearest neighbors thangalaxies that lie deep in the volume-limited samples. This isbecause potentially closer neighbors have not been observed/included in the sample. These galaxies have a higher probabilityof being selected as void galaxies than the galaxies inside thesurvey. We correct for this bias in the following way. We gener-ate a random catalog with the same angular and distance limitsas the corresponding volume-limited sample (SDSS and UZC+SSRS2) but with no clustering. We count how many randompoints lie around each of the magnitude-limited SDSS galax-ies. If the density around a galaxy is less than a certain value,we reject it from the SDSS samples. This is explained furtherbelow.

We count how many random points (N ) lie in a sphere ofsize r ¼ 3:5 h�1 Mpc around each galaxy and compute thenumber density, � ¼ N=V , where V ¼ 4�=3ð Þr3½ �. Since weknow the number of random points, the solid angle, and thedepth of the SDSS and UZC+SSRS2 surveys, we can computethe corresponding average density of random points, �random ¼N= �=3ð Þr3. Galaxies with values of � < �random are rejected, asit is their proximity to the sample’s boundaries that causes a lowvalue of �.

We apply the above procedure twice: once when we com-pare the distant SDSSmagnitude-limited sample with the SDSSrandom catalog and again when we compare the nearby SDSSmagnitude-limited sample with the UZC+SSRS2 random cat-alog. The distant SDSS sample is reduced from 65,186 galaxiesto 13,742 galaxies; the nearby SDSS sample is reduced from3784 galaxies to 2450 galaxies. The nearby SDSS sample is cutless drastically, as the UZC+SSRS2 sample covers a greaterarea.

Because the SDSS is not finished, the angular selectionfunction is complicated (see Fig. 1). An algorithm to quantifythe fraction of galaxies that have been observed in any givenregion, i.e., the completeness, has been developed and is de-scribed in Tegmark et al. (2002). The completeness for anygiven (� , � ) coordinate is returned, allowing a random cata-log with the same angular selection function to be created. Forthe SDSS, the completeness within the regions that have beenobserved is typically greater than 90%. The angular selectionfunction for the combined UZC+SSRS2 sample is easier, asthe surveys are finished and the completeness for the UZC is�96%.

3.2. Nearest Neigghbor

We classify galaxies that have a large distance to their nthnearest neighbor as void galaxies. We follow the work of El-Ad & Piran (1997) and Hoyle & Vogeley (2002) and use n ¼ 3rather than n ¼ 1 in the nearest neighbor analysis. Becausegalaxies are clustered, it is not unreasonable to expect that somegalaxies in large-scale voids might be found in binaries ortriplets. If we used n ¼ 1, then a pair of galaxies in an otherwiselow-density environment would not be classified as void gal-axies. Setting n ¼ 3 allows for a couple of bright neighbors butexcludes galaxies in typical groups. Note that we do not makeany corrections for peculiar velocities along the line of sight.

Therefore, we might underestimate the density of systems withlarge-velocity dispersions, which may pollute the void galaxypopulation. This effect could lead us to slightly underestimatethe differences between the void and wall populations.

To identify void galaxies in the SDSS, we compute the dis-tance from each galaxy in the apparent magnitude–limitedsample to the third nearest neighbor in a volume-limited sam-ple. In other words, the volume-limited sample is used to definethe galaxy density field that traces voids and other structures.We compute the average distance to the third nearest neighbor,d(3)sep, and the standard deviation, �

(3)sep, of this distance.We fix the

critical distance dcrit to be 7 h�1 Mpc, which is approximatelyequal to d(3)sep þ 1:5 � (3)

sep found from the two samples (the actualvalues are given in xx 3.3 and 3.4). This threshold is consis-tent with the criterion for defining wall and void galaxies inVOIDFINDER (Hoyle & Vogeley 2002). Galaxies in the ap-parent magnitude–limited sample whose third nearest neigh-bor lies farther than �dcrit ¼ 7 h�1 Mpc are classified as voidgalaxies. We thereby divide the apparent magnitude–limitedSDSS sample into two mutually exclusive subsamples, whichwe hereafter refer to as void and wall galaxies.

Note that the boundary of the void and wall samples is de-fined by throwing away galaxies that lie within 3.5 h�1 Mpc ofthe survey edge, whereas galaxies are classified as void galaxiesif they have less than three neighbors in a sphere of 7 h�1 Mpc.If we use 7 h�1 Mpc to mark the boundary, then the volumeavailable for finding void galaxies is decreased, especially at thenear edge of the distant sample. We tolerate this inconsistencyin order to have overlap between the near and distant samples interms of the magnitude ranges that each sample probes. How-ever, this means that there is a slightly higher probability ofa galaxy being flagged as a void galaxy near the edges, ratherthan deep in the survey. To test what effect this has, we con-struct 10 mock volume- and flux-limited catalogs from the VirgoConsortium’s Hubble Volume z ¼ 0 �CDM simulation (Frenket al. 2000; Evrard et al. 2002) that have the same geometry asregion 2. Following the procedure used in the survey data, wethrow away galaxies in the flux-limited mock catalogs that liewithin 3.5 h�1 Mpc of the region’s edge and then find the mockvoid galaxies. We then compute the average N (r) distribution ofthe mock void and wall samples, where N (r) is the normalizednumber of galaxies as a function of comoving distance, and showthese in Figure 2 (left). It can be seen that within the errors the twodistributions are similar. The exception is at the near edge wheremore mock galaxies are classified as void galaxies, as expected.Out to 125 h�1 Mpc there are 50% more void galaxies than wallgalaxies. This excess is only 4% of the whole void sample be-cause most of the void galaxies are found at greater distances.The N (r) plot for the data (Fig. 2, right) indicates a somewhatlater ratio of void/wall galaxies at the near edge of the sample.From the test above, only 4% of the void galaxies might beerroneously flagged. The rest of the difference is due to large-scale structure within the volume surveyed.

Thus we conclude that our procedure for identifying voidgalaxies and removing objects (both void and wall) near thesurvey boundaries does not produce any significant bias in theredshift distribution of void and wall galaxies. The difference isinsufficient to generate the large observed differences betweenvoid and wall galaxies. In fact, dilution of the void galaxy sam-ple can only decrease the apparent statistical significance ofdifferences between the void and wall galaxy populations; i.e.,the true differences between void and wall galaxies may bemore severe than we find. The converse is not possible; this

PHOTOMETRIC PROPERTIES OF SDSS VOID GALAXIES 53No. 1, 2004

dilution could not cause the population differences that weobserve. In x 5 we discuss the impact of this dilution on ourresults. Also, tests with smaller ‘‘clean’’ samples show, as ex-pected, a higher statistical significance, and results from thedistant sample and the nearby sample, which suffers less fromthis effect because of a wider opening angle, show the sametrends, which is further evidence that the dilution of the distantsample is a minor effect.

To test that our procedure identifies genuine void galaxies,we compute the mean, median, and upper bound of the densitycontrast (��=�) around void galaxies and compare these val-ues to the emptiness of voids as defined by VOIDFINDER.The number of galaxies in the SDSS volume-limited sampleis 22,866, and the respective volume V ¼ �=3ð Þðr3 � r30Þ¼0:6=3ð Þð2603 � 1003Þ¼ 3:32 ;106 h�3 Mpc3; therefore, themean density is � ¼ 6:84 ;10�3 h3 Mpc�3. The void galaxiescontain less than three neighbors in a sphere of 7 h�1 Mpc; thus,the density around the void galaxies is �v � 4=(4�=3)73 ¼2:78 ; 10�3 h3 Mpc�3. Therefore, the density contrast aroundvoid galaxies in the distant sample is (�v � �)=� � �0:6. Thisnumber is very similar for the nearby sample. It is an upperbound, as the median third nearest neighbor distance to the voidgalaxies is closer to 8 h�1 Mpc, giving values for the densitycontrast closer to ��=� ¼ �0:8. This value is low, although notas low as that found by VOIDFINDER for the density contrastof the voids. Since we are centered on a galaxy and galaxies areclustered, we expect the density around void galaxies (�vg) tobe higher than the mean density of a void (�v). Recall that themean density of a void is about 0.1 times the mean density ofthe universe (�universe), and since the correlation length (�) onspheres of 8 h�1 Mpc is �1 (�8 � 1), then �vg(r � 7 h�1

Mpc) ¼ �v 1þ �(r ¼ 7 h�1 Mpc)½ �� 2 ; �v. In addition, voidgalaxies are typically found near the the edge of the void wherethe density is higher.

It is important to keep in mind that since most of the voidgalaxies will lie near the edges of voids, the typical density

contrast around void galaxies is less extreme than the densitycontrast of the whole void region (see Fig. 11 in Benson et al.2003). The average number of volume-limited galaxies in asphere of 7 h�1 Mpc around a wall galaxy is 25 compared to 2around a void galaxy, demonstrating that void galaxies reallyare in highly underdense regions.

3.3. Distant Void Galaxies

For SDSS galaxies that lie in the distant SDSS sample,we use the SDSS volume-limited sample to define the galaxydensity field. Using the third nearest neighbor (n ¼ 3), weobtain (d (3)

sep; �(3)sep) ¼ (3:6; 2:10) h�1 Mpc, from which we ob-

tain dcrit ¼ 6:75 h�1 Mpc, which we round up to 7 h�1 Mpc.From the distant SDSS sample of 13,742 galaxies we find 1010void galaxies. This sample of void galaxies is referred to as voidgalaxy distant (VGD). The sample of 12,732 nonvoid galaxieswe label wall galaxy distant (WGD). The fraction of void gal-axies in the distant sample is �8%. This is only slightly higherthan the fraction of void galaxies found by VOIDFINDER(Hoyle & Vogeley 2002) and by El-Ad & Piran (1997).Figure 3 shows a redshift cone diagram of the SDSS wall

galaxies (larger dots) and the corresponding void galaxies,VGD (small dots). We plot only galaxies with j�jP15�. Notethat some of the void galaxies appear to be close to wall gal-axies. This is merely a projection effect. All the void galaxieshave less than three neighbors within a radius of 7 h�1 Mpc.After obtaining the WGD and VGD samples, we split each

void and corresponding wall galaxy sample into approximatelyequal halves by applying an absolute magnitude cut. In thiscase, the magnitude cut is done atMr ¼ �19:5, from which weobtained the corresponding subsamples, [WGD_b, VGD_b](Mr � �19:5, b = bright) and [WGD_f, VGD_f ] (Mr > �19:5,f = faint). The approximate range of absolute magnitudescovered by the subsamples is�22 P Mr � �19:5 for the brighthalf and �19:5 < Mr � �17:77 for the faint half. Figure 4shows the distribution of absolute magnitudes for the distant

Fig. 2.—N (r) distribution of void and wall galaxies from mock catalogs (left) and data (right). The left plot shows the N (r) distribution as a function of comovingdistance for the void (solid line) and wall (dotted line) mock samples. The mock samples are averaged over 10 independent realizations of region 2 (equatorial stripein the NGC; see details in x 2.1) from the Virgo Consortium’s Hubble Volume z ¼ 0 �CDM simulation (Frenk et al. 2000). The error bars are the 1 � errors on themock void galaxy samples. The right plot shows the same distribution for the distant void (solid line) and wall (dotted line) galaxy samples.


samples. Note that the terms ‘‘bright’’ and ‘‘faint’’ in this con-text are used to describe the subsamples relative to their parentsample.

3.4. Nearby Void Galaxies

To find faint void galaxies, which are present in theSDSS sample only at small comoving distances, we use theUZC+SSRS2 volume-limited sample to trace the voids, be-cause the slicelike SDSS samples are too thin to detect three-dimensional voids in this nearby volume.

The number of galaxies in the SDSS nearby sample, afterapplying the boundary corrections, is 2456. We measure thedistance to the third nearest UZC+SSRS2 volume-limited gal-axy and obtain the values (d (3)

sep; �(3)sep) ¼ (3:9; 1:9) h�1 Mpc;

hence the choice of dcrit ¼ 7 h�1 Mpc is still applicable. In thiscase we find 194 void galaxies. We refer to this void galaxysample as void galaxy nearby (VGN) and the respective parentwall galaxy sample (after removing the respective void galaxies)as wall galaxy nearby (WGN).

We again apply an absolute magnitude cut to the VGN andWGN samples. For the nearby sample, this cut is done at Mr ¼�17:0 (see Fig. 5). This cut divides the wall and respective voidgalaxy samples into approximately equal halves, which we la-bel [WGN_b, VGN_b] (Mr � �17:0, b = bright) and [WGN_f,VGN_f ] (Mr > �17:0, f = faint). The range of absolute mag-nitudes included in each subsample is�19:7 P Mr � �17:0 forthe bright half and �17:0 < Mr P �13:0 (see Fig. 5) for thefaint half. In this case the percent of void galaxies found is 8:6%.

4. PHOTOMETRIC PROPERTIES

To examine whether void and wall galaxies have differentphotometric properties, we compare their colors (u� r and

Fig. 3.—Redshift space distribution of void and wall galaxies. The largerdots show a cone diagram of the SDSS wall galaxies (100 h�1 Mpc < r <260 h�1 Mpc,Mr P �17:5). The smaller dots show the void galaxies from theapparent magnitude-limited sample (r < 17:5). We only plot galaxies withj�j P 15

�. Note that some of the black section appears to be close to magnitude-

limited galaxies. This is, however, just a projection effect (we suppress thez-direction), and all void galaxies have less than three neighbors within a three-dimensional radius of 7 h�1 Mpc.

Fig. 4.—Distribution of the absolute magnitudes in the distant wall (WGD,dotted line) and void (VGD, solid line) galaxy samples. The parent sample isapparent magnitude–limited at r � 17:5 and redshift-limited at z � 0:089. Wesplit both data sets at Mr ¼ �19:5 to obtain the void galaxy subsamples[VGD_b (b = bright), VGD_f (f = faint)] and wall galaxy subsamples[WGD_b, WGD_f ]. The cut atMr ¼ �19:5 divides both the void and the wallgalaxy samples into approximately equal halves. This histogram bins galaxies by�M ¼ 0:1. The range of absolute magnitudes probed by the void galaxies thatlie within this volume is �22:0 P Mr P�17:2.

Fig. 5.—Distributions of absolute magnitudes in the nearby wall (WGN,dotted line) and void (VGN, solid line) galaxy samples. The parent sample isapparent magnitude–limited at r � 17:5 and redshift-limited at z � 0:025. Bothdata sets are split atMr ¼ �17 to obtain the void galaxy subsamples [VGN_b (b =bright), VGN_f (f = faint)] and wall galaxy subsamples [WGN_b, WGN_f ].The cut at Mr ¼ �17 divides both the void and the wall samples into approxi-mately equal halves. This histogram bins galaxies by �M ¼ 0:1. The range ofabsolute magnitudes probed by the void galaxies that lie within this volume is�19:7 P Mr P�13:0.


g� r), concentration indices, and Sersic indices. We com-pare the properties of wall and void galaxies in both the dis-tant and the nearby samples. We also subdivide each sample byabsolute magnitude and compare their properties further. Thesamples compared are therefore (1) distant: bright (Mr � �19:5)[WGD_b, VGD_b], faint (Mr > �19:5) [WGD_f, VGD_f ],and full (undivided) void versus wall samples in each case, re-spectively, and (2) nearby: bright (Mr � �17:0) [WGN_b,VGN_b], faint (Mr > �17:0) [WGD_f, VGD_f ], and full(undivided) void versus wall samples in each case, respectively.

We compute the means of the distributions and also the er-ror on the mean to see if, on average, void and wall galaxieshave the same colors, concentration indices, and Sersic indices.We also use the Kolmogorov-Smirnov (K-S) test to see if thevoid and wall galaxies could be drawn from the same parentpopulation.

Tables 1 and 2 summarize the results of these tests for thenearby and distant samples, respectively. We present the resultsfor the whole sample as well as for the samples split by ab-solute magnitude. The results are considered in detail below.

4.1. Color

The existence of strong correlations of galaxy type withdensity (Postman & Geller 1984; Dressler 1980), galaxy typewith color (Strateva et al. 2001; Baldry et al. 2004), and densitywith luminosity and color (Hogg et al. 2003; Blanton et al.2003b) are well known; bright red galaxies tend to populategalaxy clusters and tend to be elliptical, while dim blue gal-axies are less clustered and tend to be more disklike. This be-havior is shown in an analysis of SDSS galaxy photometry by

Blanton et al. (2003b; see their Figs. 7 and 8), in which theyfind that the distribution of (g� r) colors at redshift 0.1 isbimodal.Of particular interest to us is the location of the void galax-

ies in color space. Because these galaxies evolve more slowlyand interact less with neighboring galaxies than their wallgalaxy counterparts, we might expect void galaxies to be dim,blue, and of low mass and to have high star formation rates(Benson et al. 2003). We consider two color indices, u� r andg� r, because g� r measures the slope of the spectrum andu� r is sensitive to the UV flux and the 40008 break. Since theu-band magnitudes can be noisy, by looking at g� r and u� rwe are able to verify that the results are consistent and notaffected by a low signal-to-noise ratio.In Tables 1 and 2, we compare the photometric properties of

the void galaxy samples to their respective wall galaxy samples.In Figures 6 and 7 we present normalized histograms of thecolor distributions. Solid lines correspond to the void galaxysamples, and the dotted lines represent the wall galaxies. In allcases (nearby, distant, and the bright and faint subsamples), wefind that the void galaxy samples are on average bluer than thecorresponding wall galaxy samples in both colors. If we look atthe full samples, we find that the mean values of the two sam-ples are significantly different. The nearby void galaxies havemean u� r and g� r colors that are at least 3 �� bluer than thewall galaxy samples. For the distant void galaxies, the differ-ences in the means are about 4 times greater than for the nearbycase.When we split the nearby sample into the bright and faint

samples, we see that it is at the faint end where there is the

TABLE 1

Nearby Sample

Property

Void a

� ��

Wall (WGN)b

� �� K-S Probability (P)

Full (�19:9 � Mr ��14:5), [NV ¼ 194, NW ¼ 2256]

g � r ........................ 0.433 0.014 0.490 0.004 0.002

u � r ........................ 1.598 0.040 1.764 0.013 0.001

r90 /r50 ....................... 2.390 0.024 2.390 0.007 0.802

n................................ 1.388 0.034 1.456 0.004 0.506

Bright (Mr ��17:0), [NV ¼ 76, NW ¼ 1071]

g � r ........................ 0.510 0.019 0.549 0.006 0.207

u � r ........................ 1.810 0.063 1.930 0.019 0.104

r90 /r50 ....................... 2.429 0.044 2.421 0.011 0.424

n................................ 1.518 0.060 1.626 0.004 0.605

Faint (Mr >�17:0), (NV ¼ 118, NW ¼ 1185)

g � r ........................ 0.383 0.018 0.436 0.006 0.003

u � r ........................ 1.459 0.047 1.611 0.016 0.018

r90 /r50 ....................... 2.366 0.027 2.361 0.008 0.918

n................................ 1.305 0.039 1.304 0.006 0.509

Notes.—Means, errors on the means, and K-S test probabilities that the void and wallgalaxies are drawn from the same parent population for the photometric properties of void andwall galaxies in the nearby sample (r < 72 h�1 Mpc). The number of galaxies (void and wall )in each sample and subsample are listed next to the magnitude range heading as [NV (void ),NW (wall)]. Small values of P correspond to a low probability that the two samples are drawnfrom the same parent population. The K-S test shows that void galaxies appear to havedifferent colors than wall galaxies. The void galaxies appear bluer than the respective wallgalaxies in all cases, where the average difference between the means of the colors is about2 �� . However, the concentration and Sersic indices are not significantly different.

a VGN for full, VGN_b for bright, VGN_f for faint.b WGN for full, WGN_b for bright, WGN_f for faint.


greatest difference between void and wall galaxies. The sig-nificance of the K-S test is reduced because of the smallernumber of galaxies in each sample. The nearby bright andfaint void galaxies are at least 2 �� bluer than the wall galax-ies. The differences between the nearby void and wall galax-ies are not as pronounced as in the distant samples becausewe are shot-noise–limited by how many clusters there are inthe small nearby volume. In the distant sample it is very un-likely that the wall and void galaxies in both the bright and thefaint subsamples are drawn from the same parent population(P < 10�4).

We assess the statistical significance of differences in thecolor distributions using a K-S test (the values of P, the prob-ability that the two samples are drawn from the same parentpopulation, are given in the last column of Tables 1 and 2).The probability that the void and wall galaxies are drawn fromthe same parent population is low: P < 0:002 in the nearby caseand P < 10�4 in the distant case. In the distant sample it is veryunlikely that the wall and void galaxies in both the bright andthe faint subsamples are drawn from the same parent population(P < 10�4).

4.2. Concentration Index

To compare morphological properties of void and wallgalaxies, we examine the distribution of concentration indi-ces measured by the SDSS photometric pipeline (Lupton et al.1999, 2001; Stoughton et al. 2002; Pier et al. 2003). The con-centration index (CI) is defined by the ratio CI � r90=r50, wherer90 and r50 correspond to the radii at which the integrated fluxes

are equal to 90% and 50% of the Petrosian flux, respectively.A large value of CI corresponds to a relatively diffuse galaxyand a small value of CI to a highly concentrated galaxy. The CIhas been shown to correlate well with galaxy type (Stratevaet al. 2001; Shimasaku et al. 2001). Spiral galaxies are usuallyfound to have small CIs (P2.5), whereas elliptical galaxies havelarger CIs (k2.5). This bimodal behavior of the CI can beclearly seen in Strateva et al. (2001; see Fig. 8).

Figure 8 shows histograms of CI for void and wall galax-ies for both the nearby and the distant samples along with therespective bright and faint subsamples. Tables 1 and 2 showthe mean, error on the mean, and the K-S statistic found whencomparing the wall and void galaxies.

In the nearby samples, the void and wall galaxies are notdistinguished by this morphological parameter. In Table 1, wefind that the mean values of CI are very similar. The probabilitythat the distributions of CIs of void and wall subsamples aredrawn from the same parent population approaches unity and0.5 for the faint and bright subsamples, respectively. The toprow of plots in Figure 8 shows that there is indeed little dif-ference between the distributions of CI for the void and wallgalaxies in these samples.

We find that void galaxies have, on average, significantlysmaller CIs in the bright half of the distant samples. There aremore wall than void galaxies at large values of CI (CIk 2:5).The means differ by more than 7 ��. In Figure 8, in the bottomrow, all three dotted curves show this behavior. In Table 2, it isclear that in the full sample and the bright sample, the wall andvoid galaxies have significantly different CI distributions. A

TABLE 2

Distant Sample

Property

Void a

� ��

Wall b

� �� K-S Probability (P)

Full (�22:5 � Mr ��17:77), [NV ¼ 1010, NW ¼ 12732]

g � r ................................. 0.615 0.007 0.719 0.002 <10�4

u � r ................................. 1.958 0.018 2.219 0.005 <10�4

r90 /r50 ................................ 2.449 0.011 2.571 0.004 <10�4

n......................................... 1.718 0.024 2.051 0.002 <10�4

Bright (Mr ��19:5), [NV ¼ 409, NW ¼ 7831]

g � r ................................. 0.686 0.009 0.765 0.002 <10�4

u � r ................................. 2.126 0.026 2.343 0.006 <10�4

r90 /r50 ................................ 2.505 0.019 2.656 0.004 <10�4

n......................................... 1.908 0.042 2.285 0.003 <10�4

Faint (Mr > �19:5), [NV ¼ 601, NW ¼ 4901]

g � r ................................. 0.567 0.009 0.645 0.003 <10�4

u � r ................................. 1.844 0.024 2.020 0.009 <10�4

r90 /r50 ................................ 2.411 0.013 2.435 0.005 0.211

n......................................... 1.589 0.028 1.677 0.003 <10�4

Notes.—Means, errors on the means, and K-S test probabilities that the void and wall galaxiesare drawn from the same parent population for the photometric properties of void and wall galaxiesin the distant sample (100 h�1 Mpc � r � 260 h�1 Mpc). The number of galaxies (void and wall)in each sample and subsample are listed next to the magnitude range heading as [NV (void),NW (wall)]. Small values of P correspond to a low probability that the two samples are drawn fromthe same parent population. In this case, the K-S test shows that the void and wall galaxies are drawnfrom different populations based on both color and morphology (CI and SBP). The differencesbetween the means of the different parameters measured are on average greater than 5 ��, except forthe CI in the faint subsample, where the difference is�2 ��. Void galaxies are on average bluer andmore disklike than wall galaxies.

a VGD for full, VGD_b for bright, VGD_f for faint.b WGD for full, WGD_b for bright, WGD_f for faint.


K-S analysis of the bright subsample reveals that there is aprobability of less than 10�4 that the void and wall galaxies aredrawn from the same parent population. In the case of the distantfaint void and wall galaxy samples the results are consistent.

4.3. Sersic Index

As another measure of morphology of void and wall gal-axies, we examine the Sersic index (Sersic 1968), found byfitting the functional form I(r) ¼ I0 exp (�r1=n), where n is theSersic index itself, to each galaxy surface brightness profile(SBP). With this form, n ¼ 1 corresponds to a purely expo-nential profile, while n ¼ 4 is a de Vaucouleurs profile. We usethe Sersic indices as measured by Blanton et al. (2003b) for theSDSS galaxies.

In Figure 9, we plot histograms of Sersic indices measuredfor all the samples. Statistics of these distributions and the re-sults of comparison of void and wall subsamples are listed inTables 1 and 2. In the nearby survey volume, we find n P1:5 forall void and wall galaxy subsamples, and there are no statis-tically significant differences between the distributions. The

top panels of Figure 9 show histograms of the Sersic index;the distributions of the void (solid lines) and wall (dotted line)galaxies appear very similar.We find significant differences between void and wall gal-

axies in the distant samples. The bottom panels in Figure 9show the distribution of Sersic indices for the void and wallgalaxies. For the more distant void galaxies, we find in Table 2that ndistant ¼ 1:7, which is higher than what was found for thenearby void galaxies (nnearby ¼ 1:4). A K-S test reveals thatthe void galaxies are distinct from the wall galaxies. There is aprobability of P < 10�4 in the fainter (Mr > �19:5), brighter(Mr < �20:3), and full samples that the void and wall galax-ies are drawn from the same parent population. The means ofthe Sersic indices of the void and wall galaxies differ by at least3 �� .

5. DISCUSSION

The above analysis clearly shows that there is a differencein the photometric properties of void and wall galaxies. Voidgalaxies are fainter and bluer than wall galaxies in all cases.

Fig. 6.—Color distributions of nearby void galaxies (solid lines) compared to the nearby wall galaxies (dotted lines) in two optical colors, u� r (top row) and g� r(bottom row). The first , second, and third columns are the undivided (full), bright (Mr � �17), and faint (�17 < Mr) samples, respectively. The fraction of galaxies per0.05 bin of color is shown on the Y-axis. In all cases, the solid curves are shifted to the left ; i.e., on average, the void galaxies are bluer than wall galaxies (see Table 1). AK-S test reveals that in the case of the faint subsample and full sample it is very unlikely (P P 0:007) that the full void and wall galaxy samples are drawn from the samerespective parent populations. In the bright subsample, we see an excess of luminous red galaxies at g� r � 0:9 that is not present in the void galaxy histogram.


Previous observational studies (e.g., Vennik et al. 1996; Pustilniket al. 2002; Popescu et al. 1997) suggested that isolated galax-ies in voids can be distinguished from nonvoid galaxies based ontheir color with a large enough sample. Here we provide sucha sample and even extend the analysis to compare subsamplesof void and nonvoid galaxies of similar luminosity and SBP.

Nearby, the question might be raised as to whether it is thefaintest void galaxies that are particularly blue and whetherthese galaxies dominate the statistics. To test for this, the nearbysample is cut at �17.0, reducing the range of absolute magni-tude in each bin, and again the void galaxies are bluer than thewall galaxies. A further test was made in which the galaxieswere divided into bins of �M ¼ 1 mag, and still the void gal-axies are bluer in every bin; thus the differences in color arenot dominated by the tail of the distribution. Void galaxies aregenuinely bluer than wall galaxies of the same luminosity.

In the distant sample the differences in color are only partlyexplained by the paucity of luminous red galaxies in voids. Theaverage galaxy in the distant sample has an absolute magnitudeof around �19.5, which is more than a magnitude fainter than

an M � galaxy in the r band (M � ¼ �20:80; Blanton et al.2001). Galaxies that are thought of as bright red cluster ellip-tical galaxies are typically brighter thanM�. In the full sample,the faint sample, and the bright sample (and in the �M ¼ 1 magtest), void galaxies are still bluer than wall galaxies.

In x 3.2 we noted a small excess of void galaxies near theinner boundary of the volume that encloses the distant samples.We predicted that this might affect the purity of our void galaxysamples and thereby lower the apparent statistical significanceof differences between the void and wall galaxy populations.To test for this effect, we redo selected analyses to compare thephotometric properties of void and wall galaxies in the range ofcomoving coordinate distance from r ¼ 160 to 260 h�1 Mpc,far from the region where the excess is observed near ther ¼ 100 h�1 Mpc inner boundary. We find that the differencesbetween the photometric properties of void and wall galaxiesare indeed larger for galaxies in this more restricted redshiftrange. For example, in u� r, u� g, and g� r the differencesrise to greater than 7 �� . The sense of these differences is thesame as for the larger sample; void galaxies are bluer and of

Fig. 7.—Color distributions of distant void galaxies (solid lines) compared to the distant wall galaxies (dotted lines) in two optical colors, u� r (top row) and g� r(bottom row). The left, middle, and right panels are the undivided (full), bright (Mr � �19:5), and faint (�19:5 < Mr) samples, respectively. The fraction of galaxiesper 0.05 bin of color is shown on the Y-axis. In all cases, the solid curves are shifted to the left; i.e., on average, the void galaxies are bluer thanwall galaxies (see Table 2).A K-S test reveals that it is very unlikely (P < 10�4) that the void galaxy and wall galaxy samples are drawn from the same parent populations. We clearly see an excessof luminous red galaxies in the wall galaxy histograms as a peak at g� r � 1:0.


Fig. 8.—Concentration index distribution. We show the normalized fraction of void (solid lines) and wall galaxies (dotted lines) as a function of r90=r50. The toprow shows the results for the nearby galaxies; the bottom row shows the results for the distant galaxies. The left, middle, and right panels are again the full, bright, andfaint samples. The fraction of galaxies per 0.1 bin of concentration index is shown on the Y-axis. The shift in the distribution for the distant, bright (Mr ��19:5) wallgalaxies around CI � 2:5 corresponds to the excess of bright red galaxies in the wall samples. The K-S statistic reveals that the distant void galaxy (bright and full) andrespective wall galaxy samples are very different from one another, with a probability of less than 0.01% that they are drawn from the same parent population. In thecase of the nearby galaxies, the two distributions are indistinguishable.

later type than wall galaxies. We include the more nearby, per-haps slightly diluted, void galaxy sample in our full analysisbecause it allows us to probe a larger range of absolute mag-nitude. In fact, we find consistency of results in the nearby anddistant samples over the range of absolute magnitude wherethese samples overlap. The statistical significance is compara-ble perhaps because the nearby samples are relatively smaller,albeit purer. We expect that the statistical significance of thesecomparisons will rise in future, more complete samples fromthe SDSS.

One might ask if the observed differences in color are simplythe result of the well-known morphology density relation, ex-trapolated down to lower densities: blue spiral galaxies arefound in low-density environments, while red elliptical galax-ies are found in clusters. This explanation seems unlikely in thenearby samples, where the SBPs of void and wall galaxies arequite similar. Thus, in the nearby samples, the difference incolor is not clearly linked to morphology. In the distant sam-ples, however, we see a morphological difference between thevoid and wall sample; there are more elliptical-type galaxiesin the wall sample. To test if the difference in color is causedsimply by the paucity of elliptical galaxies in voids, we divide

the distant sample by Sersic index. Blanton et al. (2003b) usen < 1:5 to represent exponential disks and n > 3 for the deVaucouleurs profiles, whereas Vennik et al. (1996) use n � 1:5for exponential law fits and n � 1:6 for early-type galaxyprofile fitting. We examine the color distributions of void andwall galaxies with a Sersic index less than 1.8 and greater than1.8 to approximately split the sample into spiral and ellipticalgalaxies. We find that the void galaxies with both n < 1:8 andn > 1:8 are bluer than the wall galaxies. In u� r and g� r andfor both n < 1:8 and n > 1:8 the void galaxies are at least 3 ��bluer than the wall galaxies, and for the n < 1:8, g� r case thedifference rises to 7 �� . Again, the samples are divided intobright and faint subsamples as well as by Sersic index. Thevoid galaxies are always bluer than the wall galaxies, althoughthe significance of the K-S test is reduced because of thesmaller number of galaxies.

Thus, void galaxies are bluer than wall galaxies even whencompared at similar SBPs and luminosities. They are alsofainter and have SBPs that more closely resemble spiral gal-axies than elliptical galaxies. These findings are consistent withpredictions of void galaxy properties from a combination ofsemianalytic modeling and N-body simulations of structure

Fig. 9.—Sersic index distribution. We plot the normalized fraction of void (solid lines) and wall galaxies (dotted lines) as a function of Sersic index. The top rowshows the results for the nearby galaxies; the bottom row shows the results for the distant galaxies. The left, middle, and right panels are again the full, bright, andfaint samples. The fraction of galaxies per 0.1 bin of Sersic index is shown on the Y-axis. We cannot distinguish the nearby void galaxies from the respective nearbywall galaxy sample based on the Sersic index. However, for the distant samples, the K-S test assigns a probability of P < 10�4 that the void galaxies are drawn fromthe same parent population as the respective wall galaxies.

PHOTOMETRIC PROPERTIES OF SDSS VOID GALAXIES 61

formation in CDM models (Benson et al. 2003). One of thereasons that void galaxies are bluer than galaxies in richer en-vironments may be that star formation is an ongoing process invoid galaxies. Galaxies in clusters and groups have their sup-ply of fresh gas cut off. Therefore, star formation is suppressedin the wall galaxies.

To illustrate the range of luminosities probed by this study,we consider which members of the LG could have been in-cluded in our samples at the distances probed by the SDSSvolume. Not only the brightest members of the LG, but also LGmembers such as M31 and M33 can be detected in the distantsample, and fainter (Mr k� 16:5) members of the LG, suchas the Large Magellanic Cloud and Small Magellanic Cloud,would be included in the nearby sample.

In the nearby sample we can detect faint dwarf elliptical(dE) galaxies, which is to be expected given that about 80%of the known galaxies in the LG are dwarfs (Sung et al. 1998;Staveley-Smith et al. 1992). It is well known that while dEgalaxies have exponential SBPs (Sandage & Bingelli 1984;Bingelli et al. 1987; Caldwell & Bothun 1987), they exhibitcolor gradients that redden outward (Jerjen et al. 2000; Vaderet al. 1988; Bremnes et al. 1998) and have a uniform colordistribution (James 1994; Sung et al. 1998). On the basis oftheir color and other properties, a fraction of the void galaxiesresemble a population of dE galaxies, which have a meang� r ¼ 0:51 (Kniazev et al. 2003). Typical dE galaxies haveSersic indices n� 1:0 and MBk� 17, consistent with oursample of nearby void galaxies (see Table 1).

6. CONCLUSIONS

Using a nearest neighbor analysis, we identify void galaxiesin the SDSS. For the first time we have a sample of �103 voidgalaxies. These void galaxies span a wide range of absolutemagnitudes,�13:5 > Mr > �22:5, which are found out to dis-tances of 260 h�1 Mpc and in regions of the universe that havedensity contrast ��=� < �0:6.

In previous studies of void galaxy properties it was sug-gested (Vennik et al. 1996; Pustilnik et al. 2002; Popescu et al.1997) that void galaxies could be distinguished from nonvoidgalaxies based on their color, and a hint of them being bluer wasobserved (Grogin & Geller 1999, 2000) from a small sample ofvoid galaxies. In this paper we present a definitive result with asample of 1204 void galaxies for which the colors, concentra-tion, and Sersic indices are compared against wall galaxies.

Void galaxies are bluer than wall galaxies of the sameintrinsic brightness and redshift distribution down to Mr ¼�13:5. We demonstrate that the difference in colors is not ex-plained by the morphology-density relation. Nearby void andwall galaxies have very similar SBPs, and still the void andwall galaxies have different colors. In the distant sample thevoids and wall galaxies have different SBPs. However, when

we divide the populations further by Sersic index, the void gal-axies are still bluer. To test that the differences in color are notdue to the choice of absolute magnitude range, we compare thecolors within narrow bins of absolute magnitude. This revealsthat void galaxies are genuinely blue and that the differencesbetween the colors are not dominated by extreme objects in thetails of the void and wall galaxy distributions.Analysis of SBPs indicates that void galaxies are of later type

than wall galaxies. Comparison of the Sersic indices betweenthe distant void and wall galaxy samples including subsampleswithin a narrow range of luminosities shows that it is very un-likely (P < 10�4) that the two samples are drawn from the sameparent population. However, based on the concentration index,it is only the bright distant void and wall galaxy samples thatdiffer significantly.Our results are in agreement with predictions from semi-

analytic models of structure formation that predict that voidgalaxies should be bluer, fainter, and have larger specific starformation rates (Benson et al. 2003). The differences in colorare probably best explained in terms of star formation. Void gal-axies are probably still undergoing star formation, whereas wallgalaxies have their supply of gas strangled as they fall intoclusters and groups.In a separate paper (Rojas et al. 2004) we discuss analysis of

the spectroscopic properties (H� , [O ii] equivalent widths, andspecific star formation rates) of our void galaxies. Work inprogress reveals that the specific star formation rate of our voidgalaxies is considerably higher, consistent with our currentfindings and predictions.

Funding for the creation and distribution of the SDSSArchive has been provided by the Alfred P. Sloan Foundation,the Participating Institutions, the National Aeronautics andSpace Administration, the National Science Foundation, the USDepartment of Energy, the Japanese Monbukagakusho, and theMax Planck Society. The SDSS4 is managed by the Astro-physical Research Consortium for the Participating Institutions.The Participating Institutions are the University of Chicago,Fermilab, the Institute for Advanced Study, the Japan Partici-pation Group, the Johns Hopkins University, Los Alamos Na-tional Laboratory, the Max-Planck-Institute for Astronomy,the Max-Planck-Institute for Astrophysics, New Mexico StateUniversity, University of Pittsburgh, Princeton University, theUnited States Naval Observatory, and the University of Wash-ington. M. S. V. acknowledges support from NSF grant AST 00-71201 and a grant from the John Templeton Foundation. Weacknowledge David Goldberg for useful conversations. Wethank the referee for valuable comments.

REFERENCES

Abazajian, K., et al. 2003, AJ, 126, 2081Alonso, M. V., da Costa, L., Latham, D., Pellegrini, P. S., & Milone, A. E.1994, AJ, 108, 1987

Babul, A., & Postman, M. 1990, ApJ, 359, 280Baldry, I. K., Glazebrook, K., Brinkmann, J., Ivezic, Z., Lupton, R. H., Nichol,R. C., & Szalay, A. S. 2004, ApJ, 600, 681

Benson, A. J., Hoyle, F., Torres, F., & Vogeley, M. S. 2003, MNRAS, 340, 160Bingelli, B. 1989, in Large-Scale Structure and Motions in the Universe, ed. M.Mezzetti (Dordrecht: Kluwer), 47

Bingelli, B., Tammann, G. A., & Sandage, A. 1987, AJ, 94, 251Blanton, M. R., Lin, H., Lupton, R. H., Maley, F. M., Young, N., Zehavi, I., &Loveday, J. 2003a, AJ, 125, 2276

Blanton, M. R., et al. 2001, AJ, 121, 2358———. 2003b, ApJ, 594, 186———. 2003c, AJ, 125, 2348Bremnes, T., Bingelli, B., & Prugniel, P. 1998, A&AS, 129, 313Caldwell, N., & Bothun, G. D. 1987, AJ, 94, 1126da Costa, L. N., et al. 1998, AJ, 116, 1Dekel, A., & Silk, J. 1986, ApJ, 303, 39Dressler, A. 1980, ApJ, 236, 351Efstathiou, G., Ellis, R. S., & Peterson, B. S. 1988, MNRAS, 232, 431Eisenstein, D. J., et al. 2001, AJ, 122, 2267El-Ad, H., & Piran, T. 1997, ApJ, 491, 421El-Ad, H., Piran, T., & da Costa, L. N. 1996, ApJ, 462, L13

4 The SDSS Web site is http://www.sdss.org.


———.El-Ad, H., Piran, T., & da Costa, L. N. 1997, MNRAS, 287, 790Evrard, A. E., et al., 2002, ApJ, 573, 7Falco, E. E., et al. 1999, PASP, 111, 438Frenk, C. S., et al. 2000, preprint (astro-ph /0007362)Fukugita, M., Ichikawa, T., Gunn, J. E., Doi, M., Shimasaku, K., & Schneider,D. P. 1996, AJ, 111, 1748

Geller, M. J., & Huchra, J. P. 1989, Science, 246, 897Grogin, N. A., & Geller, M. J. 1999, AJ, 118, 2561———. 2000, AJ, 119, 32Gunn, J. E., et al. 1998, AJ, 116, 3040Hoffman, Y., Silk, J., & Wyse, R. F. G. 1992, ApJ, 388, L13Hogg, D. W., Finkbeiner, D. P., Schlegel, D. J., & Gunn, J. E. 2001, AJ,122, 2129

Hogg, D. W., et al. 2003, ApJ, 585, L5Hoyle, F., & Vogeley, M. S. 2002, ApJ, 566, 641Huchra, J. P., Geller, M. J., & Corwin, H. 1995, ApJS, 99, 391Huchra, J. P., Geller, M. J., de Lapparent, V., & Corwin, H. 1990, ApJS,72, 433

Huchra, J. P., Vogeley, M. S., & Geller, M. J. 1999, ApJS, 121, 287James, D. A. 1994, MNRAS, 269, 176Jerjen, H., Binggeli, B., & Freeman, K. C. 2000, AJ, 119, 593Karachentsev, I. D., & Makarov, D. A. 1996, AJ, 111, 794Kirshner, R. P., Oemler, A., Jr., Schechter, P. L., & Shectman, S. A. 1981, ApJ,248, L57

Kniazev, A. Y., Grebel, E. K., Pustilnik, S. A., & Kniazeva, T. F. 2003, Astron.Nachr., 324, 81

Kuhn, B., Hopp, U., & Elaasser, H. 1997, A&A, 318, 405Lupton, R. H., Gunn, J. E., Ivezic, Z., Knapp, G. R., Kent, S., & Yasuda, N.2001, in ASP Conf. Ser. 238, Astronomical Data Analysis Software andSystems X, ed. F. R. Harnden, Jr., F. A. Primini, & H. E. Payne (SanFrancisco: ASP), 269

Lupton, R. H., Gunn, J. E., & Szalay, A. S. 1999, AJ, 118, 1406McLin, K. M., Stocke, J. T., Weymann, R. J., Penton, S. V., & Shull, J. M.2002, ApJ, 574, L115

Mo, H. J., McGaugh, S. S., & Bothun, G. D. 1994, MNRAS, 267, 129Muller, V., Arbabi-Bidgoli, S., Einasto, J., & Tucker, D. 2000, MNRAS,318, 280

Park, C., Vogeley, S. M., Geller, M. J., & Huchra, J. P. 1994, ApJ, 431, 569

Peebles, P. J. E. 2001, ApJ, 557, 495Pellegrini, P. S., da Costa, L. N., & de Carvalho, R. R. 1989, ApJ, 339, 595Pence, W. 1976, ApJ, 203, 39Pier, J. R., Munn, J. A., Hindsley, R. B., Hennessy, G. S., Kent, S. M., Lupton,R. H., & Ivezic, Z. 2003, AJ, 125, 1559

Plionis, M., & Basilakos, S. 2002, MNRAS, 330, 399Popescu, C., Hopp, U., & Elaasser, H. 1997, A&A, 325, 881Postman, M., & Geller, M. J. 1984, ApJ, 281, 95Pustilnik, S. A., Martin, J.-M., Huchtmeier, W. K., Brosch, N., Lipovetsky, V. A.,& Richter, G. M. 2002, A&A, 389, 405

Rojas, R. R., et al. 2004, ApJ, submittedSandage, A., & Binggeli, B. 1984, AJ, 89, 919Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525Scranton, R., et al. 2002, ApJ, 579, 48Sersic, J. L. 1968, Atlas de Galaxias Australes (Cordoba: Obs. Astronomico)Shectman, S. A., et al. 1996, ApJ, 470, 172Shimasaku, K., et al. 2001, AJ, 122, 1238Smith, J. A., et al. 2002, AJ, 123, 2121Staveley-Smith, L., Davies, R., & Kinman, T. D. 1992, MNRAS, 258, 334Stoughton, C., et al. 2002, AJ, 123, 485Strateva, I., et al. 2001, AJ, 122, 1861Strauss, M. A., et al. 2002, AJ, 124, 1810Sung, E., Han C., Ryden, B. S., Chun, M., & Kim, H. 1998, ApJ, 499, 140Tegmark, M., Hamilton, A. J. S., & Xu, Y. 2002, MNRAS, 335, 887Thuan, T. X., Gott, J. R., III, & Schneider, S. E. 1987, ApJ, 315, L93Vader, J. P., Vigroux, L., Lachieze-Rey, M., & Souviron, J. 1988, A&A,203, 217

Vennik, J., Hopp, U., Kovachev, B., Kuhn, B., & Elsasser, H. 1996, A&AS,117, 261

York, D. G., et al. 2000, AJ, 120, 1579Zwicky, F., & Herzog, E. 1962–1965, Catalogue of Galaxies and Clusters ofGalaxies, Vol. II–IV (Pasadena: Caltech)

Zwicky, F., Herzog, E., & Wild, P. 1961, Catalogue of Galaxies and Clusters ofGalaxies, Vol. I (Pasadena: Caltech)

Zwicky, F., Karpowicz, M., & Kowal, C. 1965, Catalogue of Galaxies andClusters of Galaxies, Vol. V (Pasadena: Caltech)

Zwicky, F., & Kowal, C. 1968, Catalogue of Galaxies and Clusters of Galaxies,Vol. VI (Pasadena: Caltech)


photometric properties of void galaxies in the …photometric properties of void galaxies in the...

Documents